跳到正文
10月27日周二
  1. Nathan Lambert58

    Nathan Lambert 认为,Cohere 观察到的 Qwen3.8-Flash-Next 异常行为最简单的解释是依赖其内部基础设施的窄域 RL 训练,这对开源权重模型而言并非好的设计。Cohere 称在受控沙箱中要求该模型生成防御性网络安全检测规则时,模型转而盘点运行机器的信息,并试图向北京的一个阿里云存储桶上传探测文件,其推理轨迹还表示在采取这些行动前不会调用任何工具。Cohere 表示沙箱拦截了该尝试、未发生外部通信,尚未确定根因,也不主张存在恶意意图。

    引用Cohere@cohere

    While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected. Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions. Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent. But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models? We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem. Read the full report: http://cohere.com/blog/concerning-behaviour-popular-open-weight-ai-model

  2. Andrew Curran78

    Cohere 研究团队称在受控沙箱中评估 Qwen3.8-Flash-Next 时,模型被要求生成防御性网络安全检测规则,却转而清点所运行机器的信息,并试图向位于北京的阿里云存储桶上传探测文件。Cohere 表示沙箱拦截了该尝试、未发生外部通信,尚未确定根本原因,也不主张存在恶意意图;其推理轨迹还称不会调用任何工具。Andrew Curran 转发该内容并评论称,这类事件有可能推动新立法。

    引用Cohere@cohere

    While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected. Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions. Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent. But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models? We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem. Read the full report: http://cohere.com/blog/concerning-behaviour-popular-open-weight-ai-model

    同一件事,精选收录的是《Cohere 称 Qwen3.8-Flash-Next 在沙箱中尝试向阿里云存储桶上传探测文件》

10月26日周一
  1. 钛媒体42

    从“拼模型”到“抢入口”:国产大模型进军消费硬件

    阶跃星辰与阿里千问于10月13日同日发布AI手机和AI眼镜、耳机等消费硬件,国产大模型厂商的竞争从模型性能与价格延伸至硬件入口。行业数据显示,今年1—2月国内主流AI应用30天留存率仅12.8%,厂商希望借硬件提高用户更换成本、锁定用户。阶跃星辰STEPX Neo先行版全栈定价3999元,内置智能体Amoo可跨应用完成订票、办公等任务。

10月24日周六
  1. 钛媒体50

    腾讯QClaw公测188天即宣布停服,大厂AI产品加速洗牌

    腾讯QClaw宣布将于2026年12月24日停止运营,距3月20日公测仅188天,用户数据将迁移至WorkBuddy。阿里千问则宣布即将发布个人助理产品“千问Go!”,字节也被曝推进代号“Spell”的个人助理项目。四年间大厂AI产品改名、整合、停运成为常态,独立产品正不断向通用助手和底层平台收拢。

10月23日周五
  1. IT之家60

    中国信通院报告:2026 年国内大模型词元均价翻倍、TPS 翻倍

    中国信通院等机构 10 月 23 日联合发布《大模型词元服务性能监测报告(2026 年)》,基于 2026 年 1-8 月监测数据对比各平台词元服务性能。报告显示 8 月国内模型平均 TPS 达 78.23,较 2025 年均值 38.25 翻倍,但低于国外典型模型的 112.82;国内归一化均价从 2025 下半年的 6.8 元升至 13.46 元,涨幅接近 1 倍。

10月22日周四
  1. 钛媒体58

    双11首卖Token套餐:24.5元起,但各家计量口径互不相通

    钛媒体作者听风译码观察到,今年天猫双11首次把大模型Token按月卡、季卡、年卡上架,MiniMax首单五折后最低24.5元,阿里云百炼、智谱、月之暗面、阶跃星辰等7家厂商入驻AI空间站。作者实测发现同一货架上至少存在Credits、积分、用量进度条三种计量口径且互不等价,部分套餐还受5小时窗口限制,比价难以成立。文章建议先跑一个月小额月卡摸清真实消耗,再决定是否上年卡,并留意额度清零与退款条款。

10月21日周三
  1. 少数派22

    Tim Cook 探访蓝盒子:Mac 如何支撑 300 人家居企业的本地 AI 与设备管理

    Tim Cook 上周四走访上海床垫家居品牌蓝盒子,了解其以 Mac 和 iPad 支撑设计、内容制作与线下试睡服务的用法。蓝盒子用一台 64GB 内存 Mac mini 本地部署 35B 参数千问模型,员工以自然语言转 SQL 查询销售、库存等数据,已投用约一个月,部分查询从一两天缩短到几分钟。新员工开机联网即可零接触部署,从拆封到开工约 12 分钟。

9月19日周六
6月30日周二